About the index
Frequently asked questions.
How the catalog is structured, what its evidence labels mean, and what the platform does today.
What is Artificial Index?
Artificial Index is a searchable encyclopedia, registry, and comparison platform for current, historical, and discontinued AI systems. It connects products, organizations, model families, releases, capabilities, benchmarks, integrations, and supporting sources.
How is the knowledge architecture organized?
Records are connected rather than stored as an isolated directory:
Organization
└─ Product / application
└─ Model / version
├─ Capabilities and modalities
├─ Access, API and licensing
├─ Benchmarks and measured evidence
├─ Integrations and regions
├─ Releases and lifecycle
└─ Sources and verification
Is the backend still only planned?
No. The project already contains PostgreSQL-compatible canonical data and migrations, Nitro application routes, structured discovery, editorial audit and source-verification queues, lifecycle monitoring, and typed knowledge-graph relationships. Specialized external search or vector infrastructure can be connected later when scale requires it; it is not presented as a completed dependency.
What do verified, partial, pending, and historical mean?
Verified means key claims have supporting public evidence. Partial means some important fields are supported while others still need review. Pending means a record is indexed but awaits deeper verification. Historical preserves lineage and context and does not imply current availability or support.
What do Status and Access mean?
Status is the system’s current public availability or release state—for example generally available, preview, waitlist, invite-only, deprecated, or discontinued. It is not a measure of quality or server uptime. Access explains the channels and conditions through which someone can use it—for example web, mobile, API, enterprise contract, self-hosting, downloadable weights, or a waitlist, together with important plan, account, regional, or organizational restrictions. API, deployment, lifecycle, and operational reliability remain separate fields.
What is the difference between a catalog profile and evidence completeness?
Catalog profile available means the system has a structured page in Artificial Index. Evidence completeness measures how many of its 21 tracked fields are verified. A profile can exist while some facts remain partial or pending; availability never turns unsupported claims into verified facts.
How are company claims separated from measured evidence?
Artificial Index keeps provider positioning separate from independently measured or source-backed evidence. Missing measurements stay missing; they are not replaced with invented scores.
Why are discontinued systems included?
Historical systems explain how today’s models, products, and techniques evolved. Lifecycle status, successors, and dates help readers distinguish archival importance from current usability.
How are duplicate news items handled?
The news scrubber normalizes URLs and titles and filters exact or near-identical copies from the same source while allowing distinct coverage of the same event.
Does Artificial Index store API keys?
No. It catalogs public integration details such as documentation, authentication type, SDKs, model identifiers, pricing references, and verification dates. Private credentials should remain with the user and provider.
How many systems are indexed?
The R20 catalog contains 1,000 indexed systems, including active, historical, and discontinued entries. Evidence depth varies by record and is shown rather than concealed.